Hacker News new | past | comments | ask | show | jobs | submit login

Why is RDF a bad thing?



* Layers upon layers of complexity: Implementing CURIs alone is a non trivial task, although all that's really needed to describe entities and attributes is 128bit UUIDs.

* There is no good build-in way for authentication and trust.

* Description Logic (the foundation of OWL) has a fundamentally prescriptive philosophy, which makes it inappropriate for most practical applications.

* No good library and tool support in general, due to the complexity.

* Blank Nodes

* No good consistency mechanism for distributed data generation, and Quads (having multiple graphs) don't properly solve this.

* Using human readable entity and attr id's leads to more bike-shedding and accidental collisions than it's worth.

* High barriers to entry.

* After years of developer disappointment the earth is pretty much salted.


What you've outlined above is a common misconception about RDF. Here's another way of looking at RDF:

An abstract data definition language for structured data represented as entity relationship graphs. Basically, the same Entity-Attribute-Value model that's existed for eons plus the addition of formalized identifiers (i.e., IRIs) for denoting entities, attributes, and values.

That's it!

RDF Schema, OWL Ontology, etc.. are simply Data Dictionaries comprising terms that describe Entity Relationship Types.

Related:

[1] Understanding Data -- https://www.slideshare.net/kidehen/understanding-29894555

[2] RDF based Linked Data, Ontologies, and KnowledgeGraph -- https://www.linkedin.com/pulse/linked-data-ontologies-knowle...


Ignoring the fact that the underlying semantics of that data description are still fundamentally misguided and broken...

What you've outlined is the common misconception by the Semantic Web community about what constitutes a living and workable software ecosystem. It's the classical architecture-astronauts fallacy of ignoring this age old truth: The medium is the message.

The "but RDF only specifies the semantics of abstract graph data, all the serialisation formats are just implementation defined incidentals" argument is at the root of why there are thousands pages of specifications and 0 high quality semantic web libraries or tools. It's like saying "no, no, no, you misunderstand JSON/XML/YAML/ASN.1 it is not a (flawed?) serialisation format, it is a data model for trees, which works on any text processing capable system", true, but a straw-man.

Developers and Ontologists flocking to Protege, OWL API and Jena, is akin to dehydrated animals flocking to the last muddy and barely drinkable water-hole. They have no choice and it's mostly sh*t.

Everybody is still editing XML files by hand, there is no maintained, complete, production ready library for any language, and most of the stuff that does exist is either the cadaverous remain of a publicly funded science grant or an obscure project developed by a lone consultant who made peddling the semantic web his livelyhood without any real community.

People have to be able to use a system, and then actually use it in order to be meaningful, theoretical what ifs are irrelevant.


"CURIs"?


A standard for shorter URI/URN identifiers: https://www.w3.org/TR/curie/

It's needed because of the "readable resource name" criterion I also mentioned. Most of the complexity in RDF is actually not in the datamodel but leaks in from other parts of the web.


As far as I'm concerned its most damning property is being all about relations but having no easy way to express properties of those relations.

It's easy to express "<something> isa <cat>", but significantly harder to express "<something> isa <cat> with probability 98%". And all attempts to express the later just look like you're trying to build an impromptu relational database in order to fit an n-ary relationship.


<person> <has_diagnosis> <diagnosis>

<diagnosis> <diagnosis_type> <ADHD>

<diagnosis> <diagnosis_probability> <high>


well done!


I don't know how much that counts but every time I had to interact with RDF data I wanted to pull my eyeballs out of my sockets pretty much for the whole thing. Also, the idea that human knowledge can be categorized seems very laughable to me - I don't know of any RDF specification or ontology that holds in practice ; too many people have too many incompatible definitions for things and there is never one better than the others.


So you are not really arguing that human knowledge cannot be categorized.

In fact you are sort of arguing the opposite: Human knowledge lends itself so well to categorization that a single ontology is impossible (if it was really hard to categorize, there would be no or few ontologies).


Yes, but that's a deeper problem with the very concept of ontologies and not with RDF itself. As long as you keep the scope limited, that is you do not try to "categorize all human knowledge ever, consistently", ontologies can be used effectively.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: